Wiribu.de is a German-language publication covering technology, gaming, artificial intelligence and smartphones for readers in Germany, Austria and Switzerland. The site runs on its own PHP/MySQL content management system, on shared hosting behind Cloudflare. Our work was not a one-off SEO audit; it was an ongoing programme to improve the site's speed, how search engines and AI assistants understand it, its editorial quality processes and its social media distribution, all within the same system.

Context and requirements

A news site's traffic depends heavily on Google and, increasingly, on AI surfaces such as ChatGPT, Perplexity, Gemini and Google AI Overviews. The initial GEO audit showed that the structured data mixed www and non-www addresses, articles used a generic Article schema instead of a news schema, author information was thin and the robots rules conflicted with some AI crawlers. In the following months, Search Console data showed that publishing pace and thin content were hurting the site's visibility. The work therefore became a phased programme covering performance, technical SEO, GEO and editorial governance together.

Solution and scope

Lighthouse 100/100

In the first optimisation round, the home page scored 100/100 in all four Lighthouse categories (Performance, Accessibility, Best Practices, SEO). To get there, critical CSS was inlined, render-blocking resources were removed, the featured image on article pages was preloaded for LCP, width and height were added to every image to prevent CLS, and responsive images were set up with a four-size srcset. On the accessibility side, ARIA labels, contrast ratios and alt texts were fixed.

Technical SEO and structured data

• All internal links and schema @id values were unified on one canonical domain, and the 301 redirects were verified.
• Articles are marked up as NewsArticle, the home page as WebSite + SearchAction and NewsMediaOrganization, and category pages as CollectionPage + BreadcrumbList.
• Multiple aspect ratios for news images, Open Graph and Twitter Card fields, ETag and Last-Modified headers, a sitemap priority hierarchy and IndexNow pings were added.
• Removed content returns 410, redirects are managed in a table, and there is a custom 404 page.

GEO: getting ready for AI search

robots.txt was rewritten with content signals based on the EU copyright directive: use in search and AI answers is allowed, model training is not, and search crawlers such as GPTBot, ClaudeBot and OAI-SearchBot are explicitly allowed. An llms.txt describing the site and its editorial principles and a dynamic llms-full.txt listing all content went live. The Impressum and editorial standards pages were brought in line with German law.

Publishing gate and E-E-A-T

To avoid leaving content quality to good intentions, server-side gates were added to the publishing flow. Before an article goes live, the system checks topic depth, source verification, author expertise and an original editor's note of at least 40 words; reusing the same note on another article is rejected. New articles are compared with published content by title and body similarity; when similarity is high, no new URL is created and updating the existing article is suggested instead. An AI Overview risk analysis also flags list and fact-question formats whose answer can be given entirely in the search result. There is a daily and weekly publishing rate limit and a mode that freezes new URLs when needed; admin overrides are written to an audit log.

Author pages were made indexable; job title, areas of expertise, credentials and verified profile links feed the Person schema from a single source. Profile links are only written to the schema once they are verified in both directions. The modification date only moves when the article body actually changes.

Content pruning

Using Search Console and content audit data, thin and overlapping articles were listed and, on the editor's decision, moved back to draft or set to 410. Every pruning round came with a backup and a rollback script.

Admin panel

Besides managing articles, pages, categories, tags, comments, media, authors and roles, the panel includes: a Search Console screen with per-article indexing checks, GA4 and Tag Manager screens, a Cloudflare cache purge log, a redirect manager, content audit, an SEO score, an AI writing assistant, ad slots and partner logos, the newsletter, and quizzes with their questions. A separate Content Quality Control role can review content without admin rights.

Technical architecture

The stack is classic PHP and MySQL, with database access through prepared statements. The server runs LiteSpeed with Cloudflare in front (Full Strict SSL, HSTS, Brotli). Security headers (X-Frame-Options, nosniff, Referrer-Policy) were added and error display is disabled in production. Scheduled publishing, indexing checks, Search Console snapshots and social media queues run as cron jobs. Changes go live only after syntax checks plus publishing-gate and SEO regression tests. The home page and article view were redesigned with light and dark themes.

Integrations

• Google: Search Console API, GA4, Tag Manager, AdSense.
• Cloudflare API: cache purging on publish and update.
• Social distribution: X (with templated images and native polls), Facebook, Instagram (posts and Stories) and a LinkedIn company page, with a draft-and-approve queue and scheduled posting.
• AI: Groq and Anthropic models for writing assistance and automatic quality checks on social images.

Outcome

Building on a technical foundation that started with full marks in all four Lighthouse categories, Wiribu.de has become a news site that is open to AI crawlers, has consistent structured data and enforces publishing quality on the server. Editors manage indexing status, content risk and social distribution from the same panel.